General Statistics of Variable r, f, m

Conclusion:

RFM Analysis

  1. Recency and Frequency: Segment these using quartiles (0.25, 0.5, 0.75), with scores from 1 to 4. More recent and more frequent purchases get higher scores.

  2. Monetary: Also use quartiles for segmentation, but additionally identify the extreme outliers. These customers will be handled separately as they represent a highly valuable segment. Scores from 1 to 5, where 5 stands for outliers.

  3. Outlier Handling: Define a threshold for outliers in the "m" dimension. Here consider a fixed monetary value 100,000, depending on the business context.

  4. Assigning Scores: After segmenting based on quartiles, assign a special score to indicate customer segments.

Segmentation Rule:

(I think I've guaranteed MECE. Apart from "5" in m_score, the other $ 4\times4\times4 = 64 $ scores will be allocated into only one category and no left out) (Also this is not a good way of showing MECE. I'll beautify this)

image.png

Question: Is this a good way of showing segments? Or treemap is better?

Synthesize New Customers